Microsoft has reportedly warned staff against "tokenmaxxing" – wastefully burning through AI tokens as a proxy for productivity – as it looks to rein in the cost of its own AI tools. According to an email seen by 404 Media, Executive Vice President Jay Parikh told employees that individual divisions would be set targets and could face restrictions, stressing that outcomes for customers and the business matter more than raw consumption figures. The move matters because it signals that even Microsoft, a major backer and host of AI infrastructure, is scrutinising the costs its own AI-first push generates, and it echoes similar rethinks happening across the industry.

The email reportedly urged staff to be mindful of consumption as they increase use of GitHub Copilot, a message given extra weight by GitHub's switch to usage-based billing, measured in "AI Credits," in June 2026. Microsoft does not appear to want an overall cut in AI use, given its ambition to be an AI-first company, but rather better value for the tokens spent. One anonymous staffer told 404 Media the memo felt like an admission that Microsoft "can't afford our own AI products," though The Register notes this is an oversimplification – the company can clearly afford them, but wants usage measured against results rather than volume alone. Microsoft declined to comment further when approached by The Register.